Accurate and precise forecasting of the Indian monsoon onset is essential for food security and hazard mitigation, with improvements in agriculture sector and early warning systems. Previous work demonstrated a good skill of UK Met Office coupled initialized global seasonal forecasting system, GloSea5-GC2 at forecasting interannual variations of the seasonal mean Indian monsoon using measures of large-scale circulation. We establish the skill of the GloSea5-GC2 in forecasting Indian monsoon onset for the 20-year hindcast period (1992-2011) using the summer hindcasts, initialized in late April-to-early May. The monsoon onset is analysed using five different objective monsoon indices, designed to examine large and local-scale features of the monsoon. Indices based on large-scale dynamic and thermodynamic indices are better at estimating monsoon onset in the model rather than local-scale dynamical and hydrological indices. GloSea5 is good at predicting category-wise monsoon onset, using early, normal or late tercile categories, one month in advance and despite systematic model errors. The forecast skill is highest over parts of central, northern and coastal India (zones of extensive agriculture) with the local-scale onset index. Monsoon onset forecast skill improves with ENSO forcing when using a large-scale circulation index, with late monsoon onset coinciding with El Niño conditions and early monsoon onset more common in La Niña years.